[C.A.T.S. CEO]

I'd like to point out that this slip up is similar to the one made by a 3rd party retailer, a Internet store (can't remember it), posted C2D MacBooks for sale BEFORE Apple did. I'd say that there is a good chance of a octo-core Mac Pro soon.

[Aron Peterson]

Originally Posted by Catfish_Man 

Link? I certainly haven't seen any news about this, although it wouldn't surprise me. The only upcoming revision I know of is the SSE4 + 45nm one.

http://www.reghardware.co.uk/2007/02...s_until_q1_08/

http://www.reghardware.co.uk/2007/03...debut_lga1366/

Now imagine a quad socket mobo with quad-core CPUs and hyperthreading. 16 physical cores + 16 virtual cores. Schweet.

[CheesePuff]

Originally Posted by C.A.T.S. CEO 

I'd like to point out that this slip up is similar to the one made by a 3rd party retailer, a Internet store (can't remember it), posted C2D MacBooks for sale BEFORE Apple did. I'd say that there is a good chance of a octo-core Mac Pro soon.

Yea, slip ups happen.

[confuzedwizard]

I think you guys are off topic. Now that some time as passed, it's safe to say that the 8-core Mac Pro's aren't coming out.

[Catfish_Man]

Originally Posted by Aron Peterson 

Intel won't launch 45nm desktop CPUs until 2008? | Reg Hardware

Intel 'Bloomfield' to intro 1,366-pin socket? | Reg Hardware

Now imagine a quad socket mobo with quad-core CPUs and hyperthreading. 16 physical cores + 16 virtual cores. Schweet.

Ehhh... kinda skeptical of The Reg, but we'll see soon enough. Agreed on the cool factor though. I would be interested to see how well the scheduler in OSX handles virtual core management. It appears to be aware of them as of 10.4, but I haven't seen anyone run tests to see if it can (for example) avoid the nasty [[thread, thread], [idle, idle]] case that you can run into with hyperthreading. Even cooler would be if it attempted to monitor the mix of instructions and schedule non-conflicting tasks to virtual core pairs, but I don't know how feasible that is.

[CIA]

Originally Posted by Cadaver 

The daisy-chain nature of FireWire makes it too useful of a technology to be completely replaced by eSATA. Use up your eSATA ports, and you'll need a whole new port-multiplying enclosure. Use up your FW800 port? Just daisy-chain on the next device.

Not saying an eSATA port wouldn't be nice, though.

Well, my single Raptor saturates a FW800 bus, so for so I can't use it for a RAID anymore. For general stuff FW400 or 800 is fine, but I need to crank out 150+MB a sec. (my highpoint 2314 eSATA external 4 drive raid pumps 225MB a sec.)

I'll take what I can get, but multilane eSATA ports would be a great addition.

8 Core machines will be announced at NAB. I'm calling it now. (tho everyone expects them at that time, so not really a suprise)

[C.A.T.S. CEO]

Originally Posted by confuzedwizard 

I think you guys are off topic. Now that some time as passed, it's safe to say that the 8-core Mac Pro's aren't coming out.

Really? I guess Apple sets a date to tell us when the new products come out...

[Northeastern292]

Originally Posted by Aron Peterson 

I use a 1.5Ghz G4 PowerBook and it does everything I need a computer to do. It can play Call of Duty 2 on high settings and edit DVCPRO HD 1080P video in real time. But if I need to render out a large project I'd take the project files across to the fastest desktop I can find. I'm a laptop man however so I'm waiting for a serious beast of a laptop before my next upgrade. The current 2.33Ghz C2Ds aren't enough. C2D @ 3Ghz and maybe I'll reach for the wallet. A quad-core laptop and I'm there.

I'm still waiting on a dual-processor MacBook Pro. (I don't mean dual-core)

[Gossamer]

Originally Posted by Northeastern292 

I'm still waiting on a dual-processor MacBook Pro. (I don't mean dual-core)

Why? I would think dual cores would be much more compact for a small size, I'm not sure about temperatures but I would assume two would run hotter than one as well. Has anyone ever made a real dual processor notebook?

[C.A.T.S. CEO]

Originally Posted by Northeastern292 

I'm still waiting on a dual-processor MacBook Pro. (I don't mean dual-core)

Thats point-less, its the same thing.

[CheesePuff]

Originally Posted by C.A.T.S. CEO 

Thats point-less, its the same thing.

Why don't you try educating yourself first on a topic you obviously know nothing about before opening your mouth next time.

[CIA]

Yay run away thread!
Dual Processor laptops, only plus I could think of is the ability to have a bus running at each processors speed, so you get lot's o' bandwidth to the chips. But, really, dual core should be ok.

[onlyone-jc]

Heh, this thread never got out of control, did it?

[Simon]

Originally Posted by Northeastern292 

I'm still waiting on a dual-processor MacBook Pro. (I don't mean dual-core)

Not gonna happen anytime soon. There's no available CPU or chipset to support that.

[Aron Peterson]

Originally Posted by Simon 

Not gonna happen anytime soon. There's no available CPU or chipset to support that.

Unless desktop CPUs and chipsets are co-opted on to a laptop mobo. A couple of companies did that back in the Pentium Pro days. Beefy thick laptops with suitcase handles.

[Eug]

Originally Posted by QuadG5Man 

For anyone who converts video on a computer, it's never fast enough. I can guarantee an 8 core mac (drool....) will not be fast enough for me, but they will be the fastest available, and an impressive machine and bump in speed. For many mac users, it's about saving time, whether that means saving money or not.

Indeed. My iMac Core 2 Duo 2.33 GHz has been running over 24 hours now, encoding a single H.264 1080p video, and yes, my software uses multiple cores effectively (according to Activity Monitor).

And I don't even do this stuff for a living. If I did do this for a living, I would definitely lust after an 8-core Mac Pro.

P.S. I'm impressed as to just how quiet this machine is. My G5 iMac was fairly loud when running at full tilt, and it just had one core.

[mduell]

Originally Posted by Aron Peterson 

Now imagine a quad socket mobo with quad-core CPUs and hyperthreading. 16 physical cores + 16 virtual cores. Schweet.

Why would you add hyperthreading to Tigerton?

[Catfish_Man]

Originally Posted by mduell 

Why would you add hyperthreading to Tigerton?

Why wouldn't you? SMT has quite a good bang/buck ratio for a number of tasks (although the Pentium 4's implementation of it was pretty awful).

[Eug]

With the memory bandwidth on those setups, would SMT help that much for most stuff on 8-core machines? Just wondering. (I know it could help for some stuff though.)

P.S. This 1080p H.264 3-pass video encode has been going for over two days now on my Core 2 Duo iMac. I don't think I'm gonna do this again too often.

[Simon]

Originally Posted by Aron Peterson 

Unless desktop CPUs and chipsets are co-opted on to a laptop mobo.

No. That won't work either because Intel desktop CPUs (Conroe and Kentsfield) aren't SMP capable. Neither is the according chipset. What you'd have to do is take their workstation-grade Xeon CPUs together with the Xeon chipset. I'll leave it to you to imagine how ridiculous it would be to try to put that into a notebook.

[Aron Peterson]

Originally Posted by Simon 

No. That won't work either because Intel desktop CPUs (Conroe and Kentsfield) aren't SMP capable. Neither is the according chipset. What you'd have to do is take their workstation-grade Xeon CPUs together with the Xeon chipset. I'll leave it to you to imagine how ridiculous it would be to try to put that into a notebook.

Or Opteron. Doesn't have to be Intel does it? Regardless, I don't see the advantage over dual-core technology which runs cooler and cheaper. Performance difference isn't noticable enough.

[Eug]

Originally Posted by Aron Peterson 

Or Opteron. Doesn't have to be Intel does it? Regardless, I don't see the advantage over dual-core technology which runs cooler and cheaper. Performance difference isn't noticable enough.

I guarantee you there will be no AMD CPUs in Macs of any sort before 2010.

It's always nice to have the option though... Something to keep Intel on its toes. Competition is good.

[schalliol]

Ok, so a few days later. What's anyone's guess for when the 8-cores will be released? My guess is before or during mid-April's NAB, if a true accidental ad was posted. Apple has released products there before, and certainly an 8-core machine would tell broadcasters that Apple means business. I need to replace my dying loaded G4/733 tower soon, and I probably won't get an 8-core machine unless they rev them all to that. However, I would expect price drops on 4-core and perhaps greater features on those models when an 8-core is released.

[phobos]

I'm waiting impatiently for the new machines to be released. It's going to replace my main machine which is, hold your breath(!!!), A 667Mhz Ti Powerbook...
The speed increase I'm going to see is probably in the x10 era....
Let's hope they'll be announced somewhere around April and later

[Eug]

re:

Originally Posted by schalliol 

Ok, so a few days later. What's anyone's guess for when the 8-cores will be released? My guess is before or during mid-April's NAB, if a true accidental ad was posted. Apple has released products there before, and certainly an 8-core machine would tell broadcasters that Apple means business. I need to replace my dying loaded G4/733 tower soon, and I probably won't get an 8-core machine unless they rev them all to that. However, I would expect price drops on 4-core and perhaps greater features on those models when an 8-core is released.

I would not be surprised to see the bulk of the line remain 4-core, with only the top end model being 8-core.

BTW, why does an 8-core model "tell broadcasters that Apple means business"?

[Aron Peterson]

Originally Posted by Eug 

BTW, why does an 8-core model "tell broadcasters that Apple means business"?

In tandem with new software coming it does. There are developers on the Windows platform who are improving fast and catching up on Final Cut Studio (eg. Sony Vegas, Avid, Adobe Premiere CS3). An 8-core Mac with Leopard and FCP6 will keep Apple ahead of the game and keep broadcasters firmly entrenched in their camp.

[anthology123]

Interesting topic. Does anyone know here if applications needed to be re-coded to take advantage of the 8 cores? If so, are these upgrades likely to be free?

[onlyone-jc]

Some do need to be re-coded, others don't. Some companies are charging independent prices for a four-core and eight-core version of their software. Your question can't really be answered specifically, because each company will have different approaches to it. Some companies will likely offer free upgrades, but then again, others won't.

onlyone-jc.

[Don Pickett]

Originally Posted by CheesePuff 

Why don't you try educating yourself first on a topic you obviously know nothing about before opening your mouth next time.

Unless you're doing things which require a lot of memory throughput there isn't much practical difference between dual-core and dual processors. In a lot of everyday tasks, assuming the software is properly threaded, dual cores are more efficient than dual processors.

[schalliol]

Anyone else want to take a guess as to the release date here? I wouldn't have expected the early errant posting without new products by now. Surely Apple doesn't give these machine specs to their store programmers or distributors until shortly before launch. If Apple can run the machines as 8 core units, it certainly makes sense to offer them.

[Catfish_Man]

Originally Posted by anthology123 

Interesting topic. Does anyone know here if applications needed to be re-coded to take advantage of the 8 cores? If so, are these upgrades likely to be free?

Data-parallel applications (a 3D renderer that can assign each block of pixels to a different processor, for example) will typically be capable of supporting a number of processors equal to the number of data elements. How efficient it will be depends the % of serial code, though.

Task-parallel (I've also seen it called Functional-parallel) applications (a game that runs audio on one thread, AI on another, and physics on another, for example) are generally harder to scale, because the threads aren't identical clones of each other, and there tend to be less different tasks than elements of data.

[Simon]

Originally Posted by Don Pickett 

Unless you're doing things which require a lot of memory throughput there isn't much practical difference between dual-core and dual processors. In a lot of everyday tasks, assuming the software is properly threaded, dual cores are more efficient than dual processors.

I'm having some trouble understanding your point. I get the impression your second sentence contradicts the first one.

[MarkLT1]

Originally Posted by Simon 

I'm having some trouble understanding your point. I get the impression your second sentence contradicts the first one.

I think what he was saying, is that there are very specific (specialized) high memory throughput applications where you will see an appreciable difference on dual processors over dual cores. But in most cases (non-high memory throughput applications) you will see no difference, or even find that dual cores are more efficient than dual processors.

I have found this to be true in my applications (large Monte Carlo simulations). I have two sets of sims- one requires a very high memory footprint/throughput, and on these, my dual Opteron system (two single core processors) seem to be more efficient (on a per-clock cycle basis) than on a Core2 Duo system. However, for the other set, which has a much lower memory throughput, there is a negligible difference between the two. Dual processors has its place (versus dual cores that is), but for anything you'd be running on a laptop, I am finding it hard to see any advantage.

[C.A.T.S. CEO]

Originally Posted by MarkLT1 

I think what he was saying, is that there are very specific (specialized) high memory throughput applications where you will see an appreciable difference on dual processors over dual cores. But in most cases (non-high memory throughput applications) you will see no difference, or even find that dual cores are more efficient than dual processors.

I have found this to be true in my applications (large Monte Carlo simulations). I have two sets of sims- one requires a very high memory footprint/throughput, and on these, my dual Opteron system (two single core processors) seem to be more efficient (on a per-clock cycle basis) than on a Core2 Duo system. However, for the other set, which has a much lower memory throughput, there is a negligible difference between the two. Dual processors has its place (versus dual cores that is), but for anything you'd be running on a laptop, I am finding it hard to see any advantage.

Agreed. Nicely explained

[Don Pickett]

Originally Posted by MarkLT1 

I think what he was saying, is that there are very specific (specialized) high memory throughput applications where you will see an appreciable difference on dual processors over dual cores. But in most cases (non-high memory throughput applications) you will see no difference, or even find that dual cores are more efficient than dual processors.

I have found this to be true in my applications (large Monte Carlo simulations). I have two sets of sims- one requires a very high memory footprint/throughput, and on these, my dual Opteron system (two single core processors) seem to be more efficient (on a per-clock cycle basis) than on a Core2 Duo system. However, for the other set, which has a much lower memory throughput, there is a negligible difference between the two. Dual processors has its place (versus dual cores that is), but for anything you'd be running on a laptop, I am finding it hard to see any advantage.

Exactly what I was saying. Two processors, each with its own dedicated bus, are more efficient when large amounts of memory need to be read into and out of the CPUs as each can access memory directly, as needed. In this scenario, dual-core chips are at a disadvantage, as they share the same bus. In a perfect world this means that each only has access to half of the stated bus bandwidth. More often than not it means that one core is sitting idle, waiting to read or write, while the other core accesses memory.

However, most applications people use day to day don't require shutting huge amounts of memory into and out of the CPU. Most apps are more CPU-dependent than memory-dependent, giving dual cores a significant advantage in day-to-day use. Surfing the web or playing games don't need much system memory.

[Simon]

I understand the difference between a single dual-core CPU and dual single-core CPUs in terms of memory bus bandwidth. I still fail to see the reasoning behind saying "there isn't much practical difference" and then claiming "In a lot of everyday tasks...dual cores are more efficient than dual processors". If dual-cores are more efficient in everyday tasks there actually is a practical difference and consumers should prefer dual-core setups.

Where I work we do a lot of high performance scientific computing. We run different types of codes on dedicated workstations (Macs and Linux PCs) and Opteron (dual-CPU) as well as Xeon (dual-core) clusters (Linux). Depending on which code we are running the dual-core or the dual-CPU boxes will show an advantage. And in other cases there is no clear winner. As you say, this has a lot to do with memory throughput. Another aspect is L2 cache. There are dual-core designs where the L2 cache is shared among both cores and other designs where each core has its own cache (as is the case with dual CPU systems). Again depending on what you're doing, one might give you a clear advantage.

That said, I agree with you that for most consumers systems dual-core CPUs will be the more efficient choice.

[P]

Dualcore has an advantage in that it's easier to read the L2 cache from the other core - it's usually a unified cache, but even if it's not (Opteron) it's at least closer than going onto the motherboard and then onto a second chip.

Dual CPUs have a higher memory bandwidth IN SOME CASES. Not nearly always - often two CPUs share one FSB. G4s do it that way, and so did all older Intel CPUs. Woodcrest, Opteron and the G5 all have separate FSBs, however, and the market is definately moving in that direction.

In general, if the dual CPUs are working in parallel on large datasets that are not shared, they will win due to higher bandwidth. If the dataset is shared, and smaller, the faster L2 cache access will give a larger boost.

[Don Pickett]

Originally Posted by Simon 

I understand the difference between a single dual-core CPU and dual single-core CPUs in terms of memory bus bandwidth. I still fail to see the reasoning behind saying "there isn't much practical difference" and then claiming "In a lot of everyday tasks...dual cores are more efficient than dual processors". If dual-cores are more efficient in everyday tasks there actually is a practical difference and consumers should prefer dual-core setups.

Perhaps I was unclear: I don't think you can say that one is superior than the other without specifying what the machine will be used for. For the kinds of tasks your average consumer uses his or her machine for, I think dual-cores are better because most of the programs such a person would use--web browsers, games, music--are cpu-bound and not memory or I/O bound. However, for things which are memory or I/O bound, like your data sets, or rendering farms or large clusters, I think that dual processors are the better choice as there is less of a memory or I/O bottleneck. And, as P stated above, my statement only applies when each processor has its own dedicated bus, and I was actually thinking about the dual core G5s versus the dual processor G5s when I replied.

I do think most consumers should prefer dual-core set ups, unless they have some specific needs which mitigate that.

[Simon]

OK, I see what you're saying. Agreed.

[Eug]

"Every new Mac features powerful dual-core or quad-core Intel processors"

[P]

The text has now been changed to Every new Mac features powerful dual-core or quad-core Intel processing, emphasis mine.

[Eug]

So not surprisingly, The Inq's article predicting 8-core Mac Pros today proved to be wrong.

I'm guessing Apple could announce new Mac Pros later this month though.

[Todd Madson]

Shrug. All I know is that ET from Tellus is the #1 machine at Seti at Home and that
is waxing every other Mac ever made out there, heck, any PC as well. His RAC is
absolutely astounding.

[Simon]

Originally Posted by Eug 

So not surprisingly, The Inq's article predicting 8-core Mac Pros today proved to be wrong.

Well, they were off by one day. The 8-core Clovertown MP is here.

I'm guessing Apple could announce new Mac Pros later this month though.

Guess you won't have to hold your breath.